Method for constructing industrial pump knowledge graph fusing failure and features

Abstract

The application discloses a kind of fusion fault and feature industrial pump knowledge graph construction method.The method comprises: first, after analyzing the characteristics of unstructured data, to the data after pre-processing, ontology construction and knowledge extraction are carried out, and industrial pump fault graph is created using neo4j graph database.Then on the basis of structured data analysis, through system mechanism, experience knowledge and feature extraction, the entity and attribute extraction of feature graph is completed;Then the related features are classified by machine learning and deep learning algorithm, to provide relationship extraction method for the creation of feature graph, and create industrial pump feature graph.Finally, the fault graph and feature graph are fused by graph fusion method, and a unified knowledge graph is formed.Two partial graphs are used as search engine, and then fault report is generated.This comprehensive diagnostic capability greatly improves the efficiency and accuracy of fault solving, and can automatically generate detailed fault diagnosis report.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rotating machinery. In particular, it relates to a method for constructing an industrial pump knowledge graph that integrates faults and features. BACKGROUND

[0002] Industrial pumps are one of the key equipment in the chemical, petroleum, power, pharmaceutical and other industries, used for transporting liquids, gases or other fluids. With the continuous improvement of industrial automation level, the running state of industrial pumps is directly related to the safe and stable operation of the entire production line. Therefore, the demand for state monitoring, fault diagnosis and preventive maintenance of industrial pumps is increasing.

[0003] Some main problems existing in the state monitoring and fault diagnosis of industrial pumps include: various monitoring systems operate independently, data is scattered, and an effective information sharing mechanism cannot be formed; expert knowledge and experience are not effectively integrated, making it difficult to pass on and develop; existing fault diagnosis systems are mostly based on single-dimensional data analysis, lacking multi-angle and multi-level information support; traditional methods have poor adaptability to complex and variable working environments, making it difficult to meet dynamic adjustment needs. At present, data acquisition and analysis mainly rely on a single sensor. Although this method can reflect the running state of the equipment to some extent, it is often difficult to accurately diagnose complex faults due to the lack of understanding of the overall equipment. Traditional fault diagnosis methods mainly include threshold-based alarm systems, signal processing-based methods, machine learning-based methods, etc. However, these methods each have limitations: threshold-based alarm systems are prone to false positives or false negatives; signal processing-based methods can provide detailed physical parameter analysis, but have limited detection capabilities for atypical faults; machine learning-based methods have good generalization ability, but perform poorly when data is insufficient or feature selection is inappropriate.

[0004] In recent years, with the development of big data, artificial intelligence and Internet of Things technologies, more and more research has begun to focus on how to use these advanced technologies to improve the management and maintenance level of industrial pumps. In particular, knowledge graphs, as a structured form of knowledge representation, are widely used in multiple fields, including but not limited to recommendation systems, natural language processing, etc. In the application scenario of industrial pumps, constructing a knowledge graph can effectively integrate equipment information, operation data, expert knowledge and other resources, providing strong support for fault diagnosis. SUMMARY

[0005] The main purpose of the present application is to overcome the problem of not intuitive and comprehensive data management in the prior art, and to provide a method for constructing an industrial pump knowledge graph that integrates faults and features.

[0006] To achieve the above-mentioned purpose, the technical solution of the present application is:

[0007] A fusion of fault and feature industrial pump knowledge graph construction method, comprising the following steps:

[0008] Step 1, analyze the data related to industrial pumps, and establish an industrial pump unstructured database.

[0009] Step 2, process the vibration signal data set, and establish an industrial pump structured database.

[0010] Step 3, manually extract the relevant content in the unstructured database in step 1, and establish an ontology model combined with expert knowledge in the field of industrial pumps.

[0011] Step 4, extract knowledge from the basic information and fault maintenance obtained in step 3, and establish fault triples.

[0012] Step 5, store the triples in step 4 into a graph database to construct an industrial pump fault graph.

[0013] Step 6, extract and summarize the data in the structured database in step 2 to establish the required nodes and node attributes.

[0014] Step 7, input the features extracted in step 6 into an intelligent algorithm for fault classification and complete relation extraction.

[0015] Step 8, summarize the entities, attributes, and relationships obtained in steps 6 and 7 to establish related triples, and then store them into a graph database to construct an industrial pump feature graph.

[0016] Step 9, further optimize the industrial pump fault graph constructed in step 5 using graph fusion methods to enhance the internal connections of the graph.

[0017] Step 10, fuse the fault graph of step 9 and the feature graph constructed in step 8 using graph fusion methods to form a unified knowledge graph, and then generate a diagnosis report.

[0018] In step 1, the specific steps include: establishing the data materials of the industrial pump unstructured database includes two parts: the first part is to collect literature materials, and sort out the basic information files of industrial pumps, mainly describing the main components of industrial pumps and the possible fault types. The second part is the related materials of specific industrial pump fault maintenance, mainly including two kinds of materials: one is to obtain part of the common faults and treatment methods of clean water pumps through network crawler, mainly involving equipment phenomena, fault causes and treatment methods; the other is to collect common faults and treatment measures of slurry pumps from the ZJ series slurry pump instruction manual of the industrial pump factory, also containing equipment phenomena, fault causes and treatment methods.

[0019] The specific steps of step 2 include collecting relevant open source data sets and dividing and sorting them into csv files, and storing them in the local hard disk as a structured database.

[0020] The specific steps of step 3 include that the entities that need to be divided in the two parts of data include the first part of "component", "fault type" and the second part of "fault equipment", "fault phenomenon", "fault reason" and "fault handling", and the "research object" node is the starting point of the related graph construction.

[0021] The specific steps of step 4 include that the basic structure of the knowledge graph is usually in the form of "knowledge triple", which is a basic unit for representing the relationship between entities. The knowledge triple is in the format of <entity, relationship, entity> to clearly express the entity relationship and complete the knowledge extraction of the centrifugal pump text data.

[0022] The specific steps of step 5 include that in the created industrial pump fault graph, the knowledge in the graph needs to be frequently searched and updated, and Neo4j graph database is selected as the carrier for building and displaying the knowledge graph for convenient and effective storage and query.

[0023] The specific steps of step 6 include summarizing and inducing the related structure and mechanism knowledge of the data set, extracting the related entity nodes. The commonly used frequency domain characteristic parameters are selected, and the divided data set is one-dimensional feature extracted; the DATSDP formula The fusion image corresponding to the signal data is generated and the address of the image is obtained. The frequency domain characteristic parameters and the image address are input into the same table file as the basic data for subsequent attribute extraction. According to the specific information of each node, the corresponding attribute data is input, and the attribute extraction of the feature graph is completed.

[0024] The specific steps of step 7 include that the recognition results of machine learning algorithm and deep learning technology are combined to build a comprehensive analysis framework. The feature vector of the whole data set is input into the machine learning algorithm, and the classification result is used as the relationship extraction information. At the same time, the classification result of DATSDP fusion image input into the convolutional neural network is also used as the relationship extraction information. According to the recognition results of machine learning algorithm and deep learning, the relationship extraction of feature graph is completed.

[0025] The specific steps of step 8 include that the obtained entities, attributes and relationships are induced to establish related triples, and Neo4j graph database is also selected as the carrier for building and displaying the knowledge graph to store the triples, and the industrial pump feature graph is constructed.

[0026] The specific steps of step 9 include: The atlas fusion method flow is, first, the related type node part between the atlases that need to be connected is traversed, the nodes with similar characteristics are identified, and the similar node pairs are formed; then, instead of directly fusing the similar node pairs into a single node, the similar relationship is established between them; in this way, not only the original node information is ensured to be complete and lossless, but also the similarity of the nodes between the atlases can be effectively reflected; in addition, the other nodes connected with the similar node pairs and their relationships remain unchanged, and finally a new atlas is formed, which contains the information of the original two atlases and enhances the connection between them. Starting from the fault atlas and the feature atlas of the industrial pump, the information atlas is further optimized by this method, and the connection between the basic information and the fault maintenance parts is enhanced.

[0027] The specific steps of step 10 include: Similarly, the atlas fusion method is used to connect the fault types of the industrial pump components with the diagnosis results of the vibration signals. In this way, the direct connection between the fault types and the vibration signal features can be established, and then the fault report can be generated.

[0028] The present application firstly analyzes the characteristics of unstructured fault data, then constructs the ontology and extracts the knowledge from the preprocessed data, creates the industrial pump fault atlas by using the neo4j graph database. Then, based on the analysis of structured data, the entities and attributes of the feature atlas are extracted through system mechanism, empirical knowledge and feature extraction; then the related features are classified by using machine learning and deep learning algorithms, and a relationship extraction method is provided for the creation of the feature atlas of the industrial pump. Finally, the fault atlas and the feature atlas are fused by using the atlas fusion method to form a unified knowledge atlas. The two atlas parts are used as search engines to generate fault reports. This comprehensive diagnostic capability greatly improves the efficiency and accuracy of fault solving, and can automatically generate detailed fault diagnosis reports, helping technicians better understand the problem and take appropriate measures to successfully complete the entire diagnostic process. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0030] Figure 1 The fusion fault and feature industrial pump knowledge atlas construction method flowchart provided by the embodiment of the present application;

[0031] Figure 2An entity schematic diagram divided by the unstructured data of the industrial pump provided by the embodiment of the present application;

[0032] Figure 3 An entity relationship schematic diagram of the industrial pump fault atlas provided by the embodiment of the present application;

[0033] Figure 4 An industrial pump fault atlas schematic diagram provided by the embodiment of the present application;

[0034] Figure 5 A construction process of the industrial pump feature atlas provided by the embodiment of the present application;

[0035] Figure 6 An industrial pump feature atlas schematic diagram provided by the embodiment of the present application;

[0036] Figure 7 A knowledge atlas fusion process schematic diagram provided by the embodiment of the present application;

[0037] Figure 8 A schematic diagram of the atlas fusion method process provided by the embodiment of the present application;

[0038] Figure 9 A part of the industrial pump knowledge atlas schematic diagram provided by the embodiment of the present application fuses the fault and the feature. DETAILED DESCRIPTION

[0039] The present application will be further illustrated below in conjunction with the drawings and specific embodiments, and it should be understood that these embodiments are only used to illustrate the present application and not used to limit the scope of the present application, and after reading the present application, the modifications of various equivalent forms of the present application by those skilled in the art all fall within the scope defined by the appended claims of the present application.

[0040] Reference Figure 1 The present application discloses a fusion fault and feature industrial pump knowledge atlas construction method, comprising the following steps:

[0041] Step 1, analyze the data related to the industrial pump, and establish an industrial pump unstructured database.

[0042] Step 2, process the vibration signal data set, and establish an industrial pump structured database.

[0043] Step 3, manually extract the related content in the unstructured database in step 1, and combine the expert knowledge in the industrial pump field to establish an ontology model.

[0044] Step 4, perform knowledge extraction on the basic information and fault maintenance obtained in step 3, and establish a fault triple.

[0045] Step 5, store the triple in step 4 into a graph database, and construct an industrial pump fault atlas.

[0046] Step 6, the data in the structured database in step 2 is correlated and the features are extracted and summarized, and the required nodes and node attributes are established.

[0047] Step 7, the features extracted in step 6 are input into the intelligent algorithm for fault classification, and the relationship extraction is completed.

[0048] Step 8, the entities, attributes, and relationships obtained in steps 6 and 7 are summarized to establish relevant triples, and then stored in a graph database to construct an industrial pump feature graph.

[0049] Step 9, the industrial pump fault graph constructed in step 5 is further optimized using graph fusion methods to enhance the internal connections of the graph.

[0050] Step 10, the fault graph of step 9 and the feature graph constructed in step 8 are also fused using graph fusion methods to form a unified knowledge graph, and then a diagnosis report is generated.

[0051] This embodiment realizes the combination of existing knowledge and real-time data by constructing an industrial pump knowledge graph that fuses faults and features. This method first proceeds from two aspects: on the one hand, based on the unstructured data such as fault records and maintenance history of industrial pumps, a fault graph is constructed, which is used as a knowledge engine; on the other hand, a feature graph is established using structured data such as sensor data and operating parameters, which is used as a data engine. Next, the knowledge engine and the data engine are reorganized and fused using graph fusion technology to form a unified knowledge graph. Finally, by traversing and analyzing this fused graph, the problem can be quickly located and a detailed fault report can be generated, thereby completing the entire fault diagnosis task.

[0052] The method is described in detail below.

[0053] Step 1, establish an industrial pump unstructured database.

[0054] The collected data includes two parts: the first part is to collect literature materials and organize industrial pump basic information files, mainly describing the main components of industrial pumps and the possible fault types they may encounter. The second part is the relevant data of specific industrial pump fault maintenance, mainly including two types of data: one is to obtain part of the common faults and treatment methods of clean water pumps through web crawlers, mainly involving equipment phenomena, fault causes and treatment methods; the other is to collect common faults and treatment measures of slurry pumps from the ZJ series slurry pump instruction manual of the industrial pump factory, also containing equipment phenomena, fault causes and treatment methods.

[0055] Step 2, establish an industrial pump structured database.

[0056] This embodiment collects relevant open source data sets, and divides and arranges them into csv files, and stores them in the local hard disk as a structured database.

[0057] Step 3, ontology construction of unstructured database.

[0058] As Figure 2 is an entity diagram of unstructured data of industrial pump. According to the two-part data structure, the entities that can be divided include the first part of “component”, “fault type” and the second part of “fault equipment”, “fault phenomenon”, “fault reason” and “fault handling”.

[0059] Step 4, knowledge extraction of unstructured data, and establishment of triples.

[0060] As Figure 3 is an entity relationship diagram of industrial pump fault atlas. The relevant entities are linked and extracted, and the relationship between the entities is extracted. The knowledge triples are clearly expressed in the format of <entity, relationship, entity>, so as to complete the knowledge extraction of centrifugal pump text data. For example, “<bearing heating, cause, bearing tile scraping is not required>”, in this example, “cause” represents the relationship type, “bearing heating” is the head entity, representing the problem or phenomenon; “bearing tile scraping is not required” is the tail entity, indicating one of the reasons for bearing heating. Such representation method simply and clearly reflects the relationship between bearing heating and its potential causes.

[0061] Step 5, store triples into graph database, and construct industrial pump fault atlas.

[0062] As Figure 4 is an industrial pump fault atlas diagram. When building a knowledge graph, the most suitable storage mode and expression method should be selected according to the characteristics and application occasions of the data. In the industrial pump knowledge graph created in this chapter, the knowledge in the graph is frequently searched and updated, in order to facilitate effective storage and query. Neo4j graph database is selected as the carrier to build and display the knowledge graph to store triples, and the industrial pump fault atlas is constructed.

[0063] As Figure 5is the construction process of industrial pump feature graph, mainly including the establishment of feature graph and the classification of information features. First, the structure, mechanism and knowledge of the data set are extracted as rule nodes; then the feature extraction of vibration data is carried out as node attributes, including the frequency domain parameters in xy direction and the DATSDP fusion image extracted in the last chapter, etc. The node attributes and rule nodes are combined to complete entity extraction; next, the rule nodes are taken as classification labels, and the two kinds of features are input into the diagnostic algorithm for classification training to complete relationship extraction; finally, the feature graph triples are formed according to the corresponding relationship of "node-relation-node", and the industrial pump feature graph is constructed through the triples.

[0064] Step 6, feature extraction of structured data, establishment of required nodes and node attributes.

[0065] The related structure and mechanism knowledge of the data set are summarized and the entity nodes related thereto are extracted. After completing the node extraction, Fourier transform and other series of operations are performed on the vibration signal data in two directions to obtain the corresponding power spectral density function, and then 8 frequency domain feature parameters in two directions are calculated to obtain one-dimensional feature extraction; the DATSDP formula The fusion image corresponding to the signal data is generated and the address of the image is obtained. The frequency domain feature parameters and image address are input into the same table file as the basic data for subsequent attribute extraction. According to the specific information of each node, the corresponding attribute data is input to complete the attribute extraction of feature graph.

[0066] Step 7, input the extracted features into intelligent algorithm for fault classification to complete relationship extraction.

[0067] The recognition results of machine learning algorithm and deep learning technology are combined to construct a comprehensive analysis framework. The feature vector of the whole data set is input into the machine learning algorithm, and the classification result is taken as the information of relationship extraction. At the same time, the classification result of DATSDP fusion image input into convolutional neural network is also taken as the information of relationship extraction. According to the recognition results of machine learning algorithm and deep learning, the relationship extraction of feature graph is completed.

[0068] Step 8, the obtained entities, attributes and relationships are summarized to establish related triples, and then stored in graph database to construct industrial pump feature graph.

[0069] As Figure 6 is the schematic diagram of industrial pump feature graph, referring to the construction of information graph, also using Neo4j graph data for storage and display, exporting triples in CSV file format, and creating feature graph using Neo4j graph database.

[0070] As Figure 7is a knowledge graph fusion process diagram, first from two aspects: on the one hand, based on the fault record of industrial pump, maintenance history and other unstructured data to construct fault graph, and take it as knowledge engine; On the other hand, using sensor data, running parameters and other structured data to establish feature graph, as data engine. Next, using graph fusion technology to recombine and fuse knowledge engine and data engine, form a unified knowledge graph. Finally, through the traversal analysis of this fused graph, the problem can be quickly located and detailed diagnosis report can be generated.

[0071] As Figure 8 is a graph fusion method flow diagram, first traverse the related type node part that needs to establish contact between graphs, identify the nodes with similar characteristics from them, and form similar node pairs; Then, not directly fuse similar node pairs into a single node, but establish similar relationship between them; In this way, not only the integrity of the original node information is ensured, but also the similarity of the nodes between the graphs is effectively reflected; In addition, other nodes and their relationships originally connected with the similar node pairs remain unchanged, eventually forming a new graph, which contains the information of the original two graphs and enhances the connection between them.

[0072] The important steps in generating similar node pairs are introduced:

[0073] (1) English-Chinese mapping

[0074] Because the node names in the two graphs use different representations, usually a mix of English and Chinese, in order to ensure the consistency and convenience of subsequent processing, a mapping dictionary of English and Chinese is needed. Through this mapping dictionary, the English name in the graph can be converted into the corresponding Chinese name, so as to realize the purpose of name standardization. This not only eliminates the obstacles caused by language differences, but also simplifies the subsequent data processing process.

[0075] (2) Tokenization and word vector generation

[0076] There are some useless words and conjunctions in the node text data, which often do not carry substantive meaning, so it is necessary to segment and denoise the text. The current mature segmentation technology includes Python's Jieba segmentation library, Java's ICTCLAS and IK segmenter, and C++'s ASPSeek, etc. In this section, Jieba segmentation library is used to achieve this goal. Jieba is a very popular Chinese segmentation tool that supports multiple segmentation strategies such as precise mode, full mode and search engine mode, and can effectively remove stop words, greatly improving the efficiency and accuracy of text processing. Through Jieba segmentation, the original text can be divided into meaningful word sequences, laying a good foundation for further text analysis.

[0077] After segmentation, the text vector representation is needed, and there are mainly two categories of text vector representation methods, the first category is one-hot encoding representation, but one-hot encoding may cause a lot of redundant sparse matrix and cause dimension disaster; the second category is distributed encoding representation, and Word2Vec is commonly used in distributed encoding representation. This section uses the distributed encoding Word2Vec model to obtain word vectors.

[0078] (3) Node similarity calculation

[0079] After generating the node name word vector, the similarity of two nodes is determined. Cosine similarity is a similarity calculation method based on vector space model, commonly used for text similarity calculation. Its calculation formula is: Where A and B are two vectors, and ||A|| and ||B|| are their lengths.

[0080] Step 9, using the graph fusion method to further optimize the fault graph.

[0081] Starting from the fault graph and feature graph of industrial pumps, this method further optimizes the fault graph, enhancing the connection between the basic information and fault maintenance in the two parts.

[0082] Step 10, fuse fault graph and feature graph to form a unified knowledge graph.

[0083] As Figure 9 is a part of the fusion of fault and feature of industrial pump knowledge graph diagram, also using the graph fusion method, the fault type of industrial pump components and the diagnosis result of vibration signal are connected. In this way, the direct connection between fault type and vibration signal features can be established, and then the diagnosis report can be generated.

Claims

1. A method for constructing an industrial pump knowledge graph that fuses faults and features, characterized in that: Step 1: Analyze data related to industrial pumps, including basic information files of industrial pumps and related data of specific industrial pump fault maintenance, and establish an unstructured database of industrial pumps; Step 2: Take the open source bearing dataset as an example, analyze the structure of the dataset and preprocess it, and establish a structured database of industrial pumps; Step 3: Manually extract related content from the unstructured database in Step 1, and combine expert knowledge in the field of industrial pumps to establish ontology models for the basic information and fault maintenance parts; Step 4: Perform knowledge extraction on the ontology obtained in Step 3 to establish triples; Step 5: Store the triples in Step 4 in a graph database to construct an industrial pump fault graph; Step 6: Extract and summarize the relevant features from the structured database in Step 2 to establish the required nodes and node attributes, including: Summarize and extract the relevant entity nodes based on the relevant structure and mechanism knowledge of the dataset; select the frequency domain feature parameters to perform one-dimensional feature extraction on the dataset; use the DATSDP formula: Generate a fusion image corresponding to the signal data and obtain the address of the image; input the frequency domain feature parameters and image address into the same table file as the basis data for subsequent attribute extraction; input the corresponding attribute data according to the specific information of each node to complete the attribute extraction of the feature graph; Step 7: Input the features extracted in Step 6 into an intelligent algorithm to perform fault classification and complete relationship extraction; Step 8: Summarize the entities, attributes, and relationships obtained in Steps 6 and 7 to establish related triples, and then store them in a graph database to construct an industrial pump feature graph; Step 9: Further optimize the industrial pump fault graph constructed in Step 5 using graph fusion methods to enhance the internal connections of the graph; Step 10: Fuse the fault graph of Step 9 and the feature graph constructed in Step 8 using the graph fusion method to form a unified knowledge graph, and then generate a diagnosis report.

2. The fusion failure and feature industrial pump knowledge graph construction method according to claim 1, characterized in that: The analysis of data related to industrial pumps and the establishment of an unstructured database of industrial pumps include: the data of the unstructured database of industrial pumps includes two parts, the first part is to collect literature data and organize the basic information files of industrial pumps, which describe the main components of industrial pumps and the possible fault types they may encounter; the second part is the related data of specific industrial pump fault maintenance, including two types of data: one is the data of common faults and treatment methods of clean water pumps obtained through web crawling, which involves equipment phenomena, fault causes and treatment methods; the other is the common faults and treatment measures of ZJ series slurry pumps collected from the slurry pump instruction manual of the industrial pump factory, which also contains equipment phenomena, fault causes and treatment methods.

3. The fusion failure and feature of industrial pump knowledge graph construction method according to claim 1, characterized in that: Processing vibration signal dataset and establishing structured database of industrial pumps includes: collecting related open source datasets and dividing and organizing them into csv files, which are stored in the local hard disk as a structured database.

4. The fusion failure and feature of industrial pump knowledge graph construction method according to claim 1, characterized in that: The related content in the unstructured database is manually extracted, and an ontology model is established by combining expert knowledge in the field of industrial pumps, including dividing the entities in the two parts of data into "component parts", "fault types" in the first part and "fault equipment", "fault phenomenon", "fault cause" and "fault handling" in the second part, and taking the "research object" node as the starting point of the related graph construction.

5. The fusion of failure and feature of industrial pump knowledge graph construction method according to claim 1, characterized in that: The obtained ontology is subjected to knowledge extraction and three-tuple establishment, including constructing the basic structure of the knowledge graph in the form of "knowledge three-tuple", which is a basic unit for representing the relationship between entities, and the knowledge three-tuple is in the format of <entity, relationship, entity> to clearly express the entity relationship and complete the knowledge extraction of the centrifugal pump text data.

6. The fusion of failure and feature of industrial pump knowledge graph construction method according to claim 1, characterized in that: The three-tuple is stored in the graph database to construct the industrial pump fault graph, including creating the industrial pump knowledge graph, retrieving and updating the knowledge in the graph, and selecting Neo4j graph database as the carrier for building and displaying the knowledge graph for convenient and effective storage and query.

7. The fusion of failure and feature of industrial pump knowledge graph construction method according to claim 1, characterized in that: The extracted features are input into the intelligent algorithm for fault classification and relationship extraction, including constructing a comprehensive analysis framework combining the recognition results of machine learning algorithms and deep learning techniques; inputting the feature vector of the overall data set into the machine learning algorithm, and taking the classification result as the relationship extraction information; at the same time, inputting the DATSDP fusion image into the classification result of the convolutional neural network as the relationship extraction information; according to the recognition results of machine learning algorithms and deep learning, the relationship extraction of the feature graph is completed.

8. The fusion of failure and feature of industrial pump knowledge graph construction method according to claim 1, characterized in that: The obtained entities, attributes and relationships are summarized to establish and store related three-tuples and construct an industrial pump feature graph, including summarizing the obtained entities, attributes and relationships to establish related three-tuples, and selecting Neo4j graph database as the carrier for building and displaying the knowledge graph to store the three-tuples and construct the industrial pump feature graph.

9. The fusion of failure and feature of industrial pump knowledge graph construction method according to claim 1, characterized in that: The constructed industrial pump fault graph is further optimized by using a graph fusion method, including the graph fusion method process, which first traverses the related type node part that needs to be connected between graphs, identifies nodes with similar characteristics from them, and forms similar node pairs; then, instead of directly fusing similar node pairs into a single node, similar relationships are established between them; this not only ensures the integrity of the original node information, but also effectively reflects the similarity of the nodes between the graphs; in addition, other nodes and their relationships connected to the similar node pair remain unchanged, and a new graph is finally formed, which contains the information of the original two graphs and enhances the connection between them; starting from the fault graph and feature graph of the industrial pump, the information graph is further optimized by this method, and the connection between the basic information and fault maintenance parts is enhanced.

10. The fusion of failure and feature of industrial pump knowledge graph construction method according to claim 1, characterized in that: The utilization of the atlas fusion method fuses the fault atlas and the feature atlas to form a unified knowledge atlas and then generates a diagnosis report, which comprises: also utilizing the atlas fusion method to link the fault types of the industrial pump components with the diagnosis results of the vibration signals; thus, the direct link between the fault types and the vibration signal features can be established, and then the diagnosis report is generated.

Images